Refbank

An open repository of conversation data from referential communication tasks

Iterated reference games have been used extensively as a paradigm to study communication, pragmatics, and convention formation. These games have varied along a number of dimensions, including design features (e.g., number of participants, communication capacity) and analytic targets (e.g., change in accuracy over trials, change in description length over trials). We introduce RefBank, an effort to harmonise iterated reference game data from multiple datasets to enable the testing of cross-study moderators as well as robust, larger-scale analyses of language.